Estimation of individual-specific progression to impending cardiovascular instability using arterial waveforms

Area of Science:

• Biomedical Engineering
• Physiology
• Machine Learning in Medicine

Background:

• Standard trauma monitoring may miss compensated internal hemorrhage until shock onset.
• Delayed recognition of internal bleeding hinders timely life-saving interventions.

Purpose of the Study:

• To test if a novel machine-learning model, the compensatory reserve index (CRI), can identify tolerance to acute volume loss before changes in stroke volume or vital signs.
• To assess CRI's ability to predict hemodynamic decompensation in individuals.

Main Methods:

• Developed a CRI model using photoplethysmogram signals to estimate stroke volume (SV) and track changes during lower body negative pressure (LBNP).
• Progressive LBNP was applied to 201 healthy subjects until hemodynamic instability.
• Validated CRI on 101 subjects, classifying them into low tolerance (LT) and high tolerance (HT) groups.

Main Results:

• CRI differentiated between LT and HT groups, reaching critical levels earlier in the LT group.
• The LT group reached CRI 0.6 at 5.27 min and 0.3 at 11.39 min, versus 7.62 min and 15.35 min for the HT group.
• Standard vital signs (heart rate, blood pressure, SV) did not distinguish between the LT and HT groups.

Conclusions:

• Machine modeling of photoplethysmogram signals can accurately trend individual progression toward hemodynamic decompensation.
• CRI offers potential for early identification of blood loss and anticipation of hemodynamic instability.
• This technology may enable timely application of life-saving interventions in trauma patients.